November 2009 - Haskell-Cafe

haddock + class instances + hidden types
by Sean McLaughlin 30 Nov '09

30 Nov '09

Hi, Is there a way to make haddock not generate instance documentation for non-exported types? Right now for instance, if there is a class A in module A, and I have a hidden type B in module B and I have instance A.A B in module B, then in the documentation for module A, the (supposedly hidden) type B shows up as one of the instances. There is no hyperlink, but I don't think it should show up there. Thanks! Sean

2 1

Parsec - separating Parsing from Lexing
by Fernando Henrique Sanches 30 Nov '09

30 Nov '09

Hello. I'm currently implementing a MicroJava compiler for a college assignment (the implemented language was defined, the implementation language was of free choice). I've sucessfully implemented the lexer using Parsec. It has the type String -> Parser [MJVal], where MJVal are all the possible tokens. However, I don't know how to implement the parser, or at least how to do it keeping it distinguished from the lexer. For example, the MicroJava grammar specifies: Program = "program" ident {ConstDecl | VarDecl | ClassDecl} "{" {MethodDecl} "}". The natural solution (for me) would be: program = do string "program" programName <- identifier ... However, I can't do this because the file is already tokenized, what I have is something like: [Program_, identifier_ "testProgram", lBrace_, ...] for the example program: program testProgram { ... How should I implement the parser separated from the lexer? That is, how should I parse Tokens instead of Strings in the "Haskell way"? Fernando Henrique Sanches

5 6

control-monad-failure and mtl
by Edward Z. Yang 30 Nov '09

30 Nov '09

Hello folks, I took advantage of Thanksgiving weekend to port my application to use Control.Monad.Failure, and learned (slightly painfully) that I still needed to pick some mechanism to instantiate my failure monads as. After the experience, I have three questions/comments: 1. Why isn't there an instance for Either in mtl? (There is one for Transformers. The error message left me very puzzled there: the docs clearly claimed the instance existed, and only a little source code diving elucidated the situation.) Copying the instance declaration from the transformers version seems to fix it. 2. I was having difficulty instantiating MonadFailure as an ErrorT for an arbitrary monad. Here is an example: {-# LANGUAGE PackageImports, FlexibleContexts #-} import "mtl" Control.Monad.Error import "mtl" Control.Monad.State import Control.Monad.Failure data MyError = MyError String instance Error MyError where strMsg = MyError type MyMonad = ErrorT MyError (State Integer) failureFunction :: MonadFailure MyError m => Integer -> m Integer failureFunction 0 = failure $ MyError "Cannot use zero" failureFunction n = return (n - 1) -- instantiate monadicFunction :: MyMonad Integer monadicFunction = failureFunction 23 Which results in the following error: failure.hs:19:18: No instance for (MonadFailure MyError (ErrorT MyError (State Integer))) arising from a use of `failureFunction' at failure.hs:19:18-35 Possible fix: add an instance declaration for (MonadFailure MyError (ErrorT MyError (State Integer))) In the expression: failureFunction 23 In the definition of `monadicFunction': monadicFunction = failureFunction 23 Which seems to contradict the documentation and source code, which states: Instances: [...] (Error e, Monad m) => MonadFailure e (ErrorT e m) How do I misunderstand? 3. In a motivating example, one of the goals of MonadFailure is to let us mix the error code of third-party modules into the generic failure mode. Control.Monad.Failure appears to give the machinery for instantiating a generic failure monad, but it doesn't have any facilities for the opposite direction: that is, marshalling a specific error form into the generic error form. Am I mistaken, and if not, would it be a welcome addition to the library? Cheers, Edward

2 2

peekCString & free memory
by El Barto 30 Nov '09

30 Nov '09

Hello, I get a segfault when I do str <- peekCString ptr free ptr return (Just str) But not when I don't free the C pointer. I also get the same behaviour with ByteString.packCString... Could you please tell me if the memory is correctly freed by GHC when I don't do it myself? And how can I specify a custom free function (i.e. xmlFree function in libxml2)? Maybe I should use a data type with two fields : the String/ByteString and the ForeignPtr to the CString? Regards

4 5

Mystery operator?
by michael rice 30 Nov '09

30 Nov '09

From: http://www.haskell.org/haskellwiki/Blow_your_mind#Polynomials -- splitting in two (alternating) -- "1234567" -> ("1357", "246") -- the lazy match with ~ is necessary for efficiency, especially enabling processing of infinite lists foldr (\a ~(x,y) -> (a:y,x)) ([],[]) This works but can't find (~) operator anywhere. Please explain or site a reference. Michael

6 10

Multi-versioning and compatibility classes.
by Jason Dusek 30 Nov '09

30 Nov '09

I have been looking at the Cabal roadmap (tickets marked for bottom) and notice that no one has mentioned multi-versioning or compatibility classes. Maybe these are bad ideas? Multi-versioning seems highly desirable. Essentially, it allows us to have the same qualified names be provided by different packages for different packages. This is desirable in the case of package versions which really are no longer the same, like Parsec 2 and Parsec 3; it also allows for the case where more than one package specifies the same qualified name because it is obvious (like `Text.JSON` or some such). It stands to reason that the "package environment" specified in the `.cabal` file is enough to tell us which package we really need to use. However, there are problems, too. Like what if I have one package that depends on `parsec-2.1.0.1` and another that depends on `parsec-2.1.0.0`? With multi-versioning, they both compile fine but then I try to use a parser from one with a combinator defined in the other: this fails to compile because: parsec-2.1.0.1:Text.ParserCombinators.Parsec.GenParser is not the same as: parsec-2.1.0.0:Text.ParserCombinators.Parsec.GenParser So now we need compatibility classes. The first package might specify a package dependency like: ~parsec-2.1.0.1 where we take the `~` to mean "requires something compatible with". The second package specifies a similar dependency: ~parsec-2.1.0.0 Then the maintainer of Parsec specifies that 2.1.0.1 is compatible with 2.1.0.0 and we can compile the second package, the one that require 2.1.0.0, with 2.1.0.1 since they are compatible and that is all it is asking for. Then we try to use these two packages together and everything works. -- Jason Dusek

1 0

Define a module in multiple files
by Ozgur Akgun 30 Nov '09

30 Nov '09

Dear Cafe, Is there a way of splitting the definition of a module into multiple files? Cheers, -- Ozgur Akgun

2 3

Re: Are there standard idioms for lazy, pure, error handling?
by John Lato 30 Nov '09

30 Nov '09

> From: Bas van Dijk <v.dijk.bas(a)gmail.com> > > On Mon, Nov 30, 2009 at 6:22 AM, John Millikin <jmillikin(a)gmail.com> wrote: >> ...I've considered two possible error handling modes... > > Regarding parsing, there's a third option: iteratees[1]. See [2] for a > motivation and description of iteratees. > I think it's interesting to note that this works with iteratees because they essentially do exactly what Malcolm suggested. That is, when an error occurs, it is referenced in the internal state of the iteratee, and processing is aborted. Iteratees are strict, but they allow for many of the same benefits of laziness, and can work well for large data sets. John

1 0

ANNOUNCE: graphviz-2999.7.0.0
by Ivan Lazar Miljenovic 30 Nov '09

30 Nov '09

I'm pleased to announce the latest version of my Haskell bindings to the Graphviz [1] suite of tools for visualising graphs. As usual, it is available from Hackage [2]. [1] http://graphviz.org/ [2] http://hackage.haskell.org/package/graphviz Changes in this release include: * Updated and extended functions to interact with the Graphviz tools. This now includes: - a better listing of available output types; - distinguishing file outputs from canvas outputs; - ability to automatically add the correct file extension to file outputs - Return any errors if calling Graphviz failed rather than just printing them to stderr * Improved Color support: - fixed ColorScheme values - explicitly named X11 colors - conversion to/from values from the colour library * Removed OrientationGraph; problems with distinguishing it when parsing from node-based Orientation values; its functionality is duplicated by Rotate. * By default, the generated Dot code is now no longer indented; you can now use the prettyPrint functions in Data.GraphViz to produce readable Dot code. * Added a testsuite; this is buildable by building with --flags=test. Numerous printing and parsing bugs were picked up with this. Plans for the next release: * Determine how to deal with encodings, since Graphviz uses UTF-8 by default; the hSetEncoding function in GHC-6.12 would be ideal, but backwards-incompatible. * Add a generic version of the types in Data.GraphViz.Types that would let people have more fine-grained control over the layout of their graphs by removing the ordering restriction used. Note that this will _not_ be the default, just an option for those that want it. -- Ivan Lazar Miljenovic Ivan.Miljenovic(a)gmail.com IvanMiljenovic.wordpress.com

1 0

License question
by jean-christophe mincke 30 Nov '09

30 Nov '09

Hello, I am writing a port to F# of some haskell standard libraries. I would like to publish them under an open source license and mention the origin of the initial code. www.haskell.org is under the simple permissive license. Does this license also cover the souce code available from that site? Is there any restriction about the license I might choose? Thak you Regards jcm

4 5